1. Soil Moisture Content Inversion Model on the Basis of Sentinel Multispectral and Radar Satellite Remote Sensing Data
- Author
-
Guo, Fei, Huang, Zugui, Su, Xiaolong, Li, Yijie, Luo, Linyu, Ba, Yalan, Zhang, Zhitao, and Yao, Yifei
- Abstract
Purpose: Soil moisture constitutes a pivotal component of the hydrological cycle, vital for drought forecasting and climate change studies. Despite the development of remote sensing technology offering a robust means for soil moisture content (SMC) monitoring, the exclusive use of a single data source for SMC monitoring commonly has problems with poor timeliness and low accuracy, particularly within vegetated regions. This study aims to improve the accuracy of SMC monitoring in vegetated areas through the fusion of radar and multispectral data and to explore the influence of soil depth on SMC inversion. Methods: Shahaoqu was selected as the study area, which has a variety of crop types. During the sampling period, the crops were flourishing. Fifteen spectral indices and fourteen radar feature parameters were extracted from Sentinel multispectral and radar data, respectively, with the variable importance projection (VIP) method used for variable selection. The SMC inversion model was built using three algorithms: Multiple Linear Regression (MLR), Back Propagation Neural Network (BPNN), and Support Vector Machine (SVM). Results: The SMC model, which had synergistically combined Sentinel radar and optical data, showed improved stability and inversion accuracy. It achieved a high adjusted coefficient of determination (R2) of up to 0.731 and a root mean square error (RMSE) of 2.2%. The inversion accuracy of the three algorithms at varying soil depths ranks as follows: 0–10 cm, 10–20 cm, and 0–20 cm, from best to worst. The machine learning models ranked by inversion accuracy from highest to lowest are SVM, BPNN, and MLR. Conclusions: This study presents an efficient method for fusing Sentinel optical and radar remote sensing data for soil moisture inversion. It can effectively invert soil moisture under different vegetation types, providing essential theoretical foundations for precision agriculture and environmental management.
- Published
- 2024
- Full Text
- View/download PDF